The invention discloses an underwater positioning method based on ranging and a neural network algorithm. The method comprises the steps that K anchor nodes and N to-be-positioned nodes are arranged in positioning water; the distance between each to-be-positioned node and K anchor nodes is measured; and positioning is carried out for two stages of offline training and online positioning. The stage of offline training comprises the steps that the distance between the n-th to-be-positioned node and all anchor nodes is measured to form a distance vector, and the distance vector is input into a three-layer nerve network which comprises a neural network input layer, a neural network hidden layer and a neural network output layer; prediction model parameters are updated through algorithm iteration; and neural network model parameters are output. The stage of online positioning comprises the steps that the current distance between the n-th to-be-positioned node and all anchor nodes is measured to form a distance vector; the distance vector is input to the neural network model parameters acquired in the stage of offline training; and the position coordinates of the to-be-positioned nodes are output to reduce ranging errors.